SUMOylation of the hepatoma-derived growth factor
negatively influences its binding to chromatin
Ketan Thakar
1
, Rainer Niedenthal
2
, Elwy Okaz
1
, Sebastian Franken
3
, Astrid Jakobs
2
,
Shivangi Gupta
1
, Sørge Kelm
1
and Frank Dietz
1
1 Department of Biochemistry, Centre for Biomolecular Interactions Bremen (CBIB), University of Bremen, Germany
2 Department of Biochemistry, Hannover Medical School, Germany
3 Department of Physiological Chemistry, Rheinische-Friedrich-Wilhelm University of Bonn, Germany
Hepatoma-derived growth factor (HDGF) is the ubiq-
uitously expressed prototype of a family of proteins
called the HDGF-related proteins (HRPs) [1]. To date,
four HRPs (HRP-1 to HRP-4) and a protein called
the lens epithelium-derived growth factor (LEDGF)
have been described [2,3]. HDGF was identified as a
25 kDa heparin binding protein, purified from condi-
tioned media of the human hepatocarcinoma cell line
Huh7 [1]. Upon overexpression, HDGF shows mito-
genic activity and this growth promoting activity
depends on its nuclear localization mediated by the
presence of two functional nuclear localization signals
(NLS) within its primary amino acid sequence [4,5].
HRPs and LEDGF share a highly conserved N-termi-
nal region of approximately 100 amino acids called the
hath region (homologous to the amino terminus of
HDGF). This region includes a PWWP domain found
in an increasing number of proteins [6,7].
Structural data available for the hath region of
HDGF and HRP-3 revealed a characteristic fold made
up of a five-stranded bbarrel followed by a-helical ele-
ments [8–10]. The PWWP domain shares similarities
Keywords
HDGF related protein (HRP); nuclear
localization; PWWP domain; SUMOylation
Correspondence
F. Dietz, Department of Biochemistry,
Centre for Biomolecular Interactions
Bremen (CBIB), University of Bremen,
Leobener Strasse im NW2, 28359 Bremen,
Germany
Fax: +49 421 218 2981
Tel: +49 421 218 4324
E-mail: fdietz@uni-bremen.de
(Received 16 May 2007, revised 7
December 2007, accepted 16 January 2008)
doi:10.1111/j.1742-4658.2008.06303.x
Hepatoma-derived growth factor is a nuclear targeted mitogen containing a
PWWP domain that mediates binding to DNA. To date, almost nothing is
known about the molecular mechanisms of the functions of hepatoma-
derived growth factor, its routes of secretion and internalization or post-
translational modifications. In the present study, we show for the first time
that hepatoma-derived growth factor is modified by the covalent attach-
ment of small ubiquitin-related modifier 1 (SUMO-1), a post-translational
modification with regulatory functions for an increasing number of pro-
teins. Using a basal SUMOylation system in Escherichia coli followed by a
MALDI-TOF-MS based peptide analysis, we identified the lysine residue
SUMOylated located in the N-terminal part of the protein adjacent to the
PWWP domain. Surprisingly, this lysine residue is not part of the consen-
sus motif described for SUMOylation. With a series of hepatoma-derived
growth factor mutants, we then confirmed that this unusual location is also
used in mammalian cells and that SUMOylation of hepatoma-derived
growth factor takes place in the nucleus. Finally, we demonstrate that
SUMOylated hepatoma-derived growth factor is not binding to chromatin,
in contrast to its unSUMOylated form. These observations potentially
provide new perspectives for a better understanding of the functions of
hepatoma-derived growth factor.
Abbreviations
DAPI, 4¢-6-diamino-2-phenylindole HCL; Dnmt, DNA methyltransferase; EGFP, enhanced green fluorescence protein; HA, hemagglutinin
epitope; hath, homologous to the amino terminus of HDGF; HDGF, hepatoma-derived growth factor; HRP, HDGF-related protein; IAA,
iodacetamide; LEDGF, lens epithelium-derived growth factor; NLS, nuclear localization signal; SUMO-1, small ubiquitin-related modifier 1;
wt, wild-type.
FEBS Journal 275 (2008) 1411–1426 ª2008 The Authors Journal compilation ª2008 FEBS 1411
with the well known Tudor and Chromo domain and,
like these domains, it has been proposed to play a role
in DNA-binding and ⁄or protein–protein interactions.
In the case of HDGF, the PWWP domain may have a
dual function in binding double-stranded DNA as well
as the glycosaminoglycan heparin [2,8–10]. DNA-bind-
ing via the PWWP domain of HDGF appears to be
specific for a region covering approximately 40 bp
found in potential target genes of HDGF [11],
although it has not been clarified whether further spec-
ificity may be mediated by the C-terminal portion of
the protein [8]. A recent study by Sue et al. [12] dem-
onstrated that dimerization of the PWWP domain by
an unusual domain-swapping leads to an increased
binding affinity for heparin. However, the physiologi-
cal role of this phenomenon is unclear.
Hepatoma-derived growth factor is secreted from
cells. The mechanism for externalization remains
unclear because HDGF, like the other HRPs, has no
obvious signal peptide. Extracellular HDGF appears
to be internalized by binding to heparan sulfate or
other mechanisms [5]. Recent studies have provided
evidence for a potential receptor specifically binding
extracellular HDGF, leading to the activation of intra-
cellular signalling cascades [13].
The expression of HDGF changes during develop-
ment, as shown for kidney, liver, heart and vascular
tissue [14–20]. In addition, recent studies have demon-
strated that HDGF is differentially expressed in the
brain [21] and also can function as a potent neuro-
trophic factor [22,23]. Furthermore, different studies
have shown that HDGF can serve as a prognostic
marker in a variety of human cancers [24–31] and that
it probably promotes angiogenesis and tumor progres-
sion [32].
Phosphorylation prediction programs have identified
HDGF as a good candidate for phosphorylation on sev-
eral serine and threonine residues, but only one mass
spectroscopy based approach has confirmed the use of
serine residues S132, S133 and S165 [33], although no
evidence for functional relevance was provided.
A post-translational modification found in several
nuclear proteins comprises the attachment of the small
ubiquitin-related modifier 1 (SUMO-1). The modifica-
tion by this 11 kDa protein is mechanistically related
to that of ubiquitin, with which it shares a high degree
of structural similarity. Like ubiquitination, SUMOyla-
tion is a dynamic process that is mediated by activat-
ing (E1), conjugating (E2) and ligating (E3) enzymes
and can be reversed by the action of SUMO specific
proteases [34,35]. Despite these similarities, the func-
tions of both modifications differ. SUMOylation of
target proteins usually occurs on lysine residues in the
context of a highly conserved recognition motif
YKxE ⁄D (where Ystands for a large hydrophobic
amino acid, K is the lysine modified, x is any amino
acid and E ⁄D are the negatively charged amino acids
glutamate or aspartate). Well documented functions of
SUMOylation are the regulation of subcellular distri-
bution, DNA repair, transcriptional regulation, stabil-
ization, RNA metabolism and cell signalling [34–38].
SUMO itself can further serve as a docking site for the
binding of other proteins containing SUMO binding
motifs [39–42].
Based on the knowledge that HDGF is a nuclear
targeted mitogen with DNA binding capacity, we
investigated whether HDGF is also modified by the
addition of SUMO-1. In the present study, we show
for the first time that HDGF serves as a template for
SUMO-1 conjugation, although it does not contain a
suitable consensus site for SUMOylation. Using a
basal SUMOylation system in Escherichia coli [43]
followed by a MALDI-based peptide analysis of the
SUMOylated HDGF wild-type (wt) and a series of
HDGF mutants, we identified an unusual SUMOyla-
tion site located in the N-terminal hath region. Fur-
thermore, we discovered that SUMOylated HDGF
does not bind to chromatin, in contrast to its
unSUMOylated form.
Results
SUMOylation of HDGF
For most SUMOylated proteins, overexpression of
the target protein together with SUMO-1 is necessary
to detect SUMOylation. When untagged HDGFwt
(apparent molecular mass ¼40 kDa) is overexpressed
together with enhanced green fluorescence protein
(EGFP)-SUMO-1 in COS-7 cells (Fig. 1A) or human
embryonic kidney cells (HEK293) (data not shown) we
use the advantage of the much higher molecular mass
shift of the EGFP-SUMO-1 SUMOylated proteins.
We detected an extra protein band reacting with a spe-
cific anti-HDGF serum in the molecular weight range
expected for EGFP-SUMO-1 conjugated HDGF
(apparent molecular weight ¼100 kDa). This observa-
tion suggests that HDGF can be modified by SUMOy-
lation. In several systems, it has been shown that
SUMOylation is a highly dynamic, reversible modifi-
cation, which is sensitve to the action of specific
isopeptidases. Since these are cysteine proteases, de-
SUMOylation can be partially blocked by lysing the
cells in the presence of N-ethylmaleimide and iodaceta-
mide (IAA) to alkylate the free SH-group in the active
centres of the isopeptidases. In agreement with this
SUMOylation of HDGF K. Thakar et al.
1412 FEBS Journal 275 (2008) 1411–1426 ª2008 The Authors Journal compilation ª2008 FEBS
phenomenon, we found that most of the suspected
HDGF-EGFP-SUMO-1 band migrating at 100 kDa is
lost if the cell lysates are incubated for 15 min on ice
in the absence of N-ethylmaleimide and IAA (Fig. 1B),
whereas almost no loss occurs if these SH-alkylating
reagents are included.
To further investigate whether the additional protein
bands are indeed due to the covalent modification of
HDGF with SUMO-1, we used C-terminally Strep-
tagged HDGF, which can be precipitated using
StrepTactinbeads. Similar to the experiment with
untagged HDGFwt, the C-terminal Strep-tag-labelled
HDGF appears to be SUMOylated (Fig. 1C) because,
in StrepTactinprecipitates, a similar upshifted
HDGF-band occurs, which reacts with antibody
against SUMO-1 (Fig. 1B). As negative controls, we
cotransfected either EGFP alone or an EGFP-SUMO-
1 chimera lacking the C-terminal di-glycine motif
(EGFP-SUMO-1DGG) required for the isopeptide
bond formation of SUMO-1 with the acceptor lysine.
Only EGFP-SUMO-1 including the terminal di-gly-
cine was covalently attached to HDGF, whereas EGFP
alone or EGFP-SUMO-1DGG was not (Fig. 1C).
To overcome problems in the detection of HDGF
mutants due to the restricted specificity of the polyclonal
anti-HDGF antiserum, in all further experiments, Strep-
tag-labelled HDGF constructs were employed, provid-
ing the possibility of using identical StrepTactin
precipitation and detection assays for all HDGF
mutants. For all these experiments, similar results were
obtained if HEK293 cells were utilized instead of COS-7
cells, supporting the notion that SUMOylation of
HDGF is not a COS-7 cell restricted modification.
Overexpression of the conjugating enzyme Ubc-9
(E2) is commonly used to obtain detectable amounts
of SUMOylated products. However, this was not
required for the SUMOylation of HDGF because
omission of the hemagglutinin epitope (HA)-Ubc-9
encoding plasmid did not reduce the level of SUMOy-
lation (Fig. 1C). These data suggested that endogenous
levels of Ubc-9 in COS-7 cells are sufficient to generate
detectable levels of SUMOylated HDGF.
SUMOylation site of HDGF in E. coli
Screening the primary amino acid sequence of HDGF
from different mammalian species (Fig. 2) using the
prediction programs for SUMOylation motifs SUMO-
plot(http://www.abgent.com/doc/sumoplot) and the
SUMOsp–SUMOylation sites prediction program
(http://bioinformatics.lcd-ustc.org/sumosp/), we found
only three motifs that weakly match the postulated
consensus motif YKxE ⁄D (Fig. 2). The highest score
by SUMO site prediction was obtained for K223 in
the motif AK
223
EE of mouse HDGF (> 0.8), which is
TKED in human and chimpanzee or AKED in bovine
HDGF (Fig. 2). Other potential motifs predicted with
low scores (0.5) are EK
148
NE and PK
167
RP. Single K
to R mutations in these three motifs and the expres-
sion of the mutants together with EGFP-SUMO-1 in
COS-7 cells did not lead to any detectable loss in
SUMOylation compared to HDGFwt (Fig. 3). This
strongly suggested that other lysine residues than these
are SUMOylated in HDGF.
Since no obvious consensus motif was found in
HDGF, we aimed to identify the SUMOylation site(s)
of HDGF by MS after tryptic digestion. In order to
obtain sufficient amounts of SUMOylated HDGF for
MALDI-TOF-MS analysis, we expressed the protein
in E. coli. Essential compounds of the SUMOylation
machinery, such as the activating (E1) and conjugating
(E2) enzymes but not the ligating enzyme (E3) [43],
were coexpressed with Strep-tagged mHDGFwt in
E. coli strain BL21 DE3. This artificial bacterial
SUMOylation system enabled us to purify high levels
of SUMOylated and unmodified HDGF from bacterial
lysates via StrepTactinprecipitation. After 2D elec-
trophoresis of the purified protein, spots representing
unmodified HDGFStrep-tag appearing at a molecular
weight of 40 kDa (Fig. 4A, spot 1) and a protein spot
appearing at a molecular weight of 63 kDa (Fig. 4A,
spot 2) were cut out, digested with trypsin and used
for MALDI-TOF-MS analysis. The protein spot at
63 kDa is only observed if HDGF is coexpressed with
the SUMOylation machinery (data not shown). In the
spectra of both protein spots, we found peptide masses
perfectly matching the expected peptides from HDGF.
However, only in spot 2 did we recognize additional
peptide masses corresponding to peptides derived from
huSUMO-1 (Fig. 4C).
Furthermore, comparison of the obtained peptide
spectra of both spots clearly showed an almost com-
plete loss of two mass peaks in the chromatogram
derived from the digest of spot 2, most probably corre-
sponding to monoSUMOylated HDGF (Fig. 4D,E).
These mass peaks perfectly match the HDGF peptides
K
80
-K
96
(K
80
GFSEGLWEINNPTVK
96
) and G
81
-K
96
(G
81
FSEGLWEINNPTVK
96
) of the hath region
(Fig. 2). This observation strongly suggested that
either K
80
or K
96
is modified by SUMOylation.
SUMOylation of HDGF at Lys
80
in mammalian
cell lines
To investigate whether K
80
or K
96
is SUMOylated in
mammalian cells as suggested by the bacterial system,
K. Thakar et al. SUMOylation of HDGF
FEBS Journal 275 (2008) 1411–1426 ª2008 The Authors Journal compilation ª2008 FEBS 1413
we generated and expressed K80R and K96R mutants
of HDGF in COS-7 cells together with EGFP-SUMO-1.
Interestingly, expression of the mutants in comparison
to HDGFwt clearly showed an almost complete loss of
the signal for SUMOylated HDGF only in the case of
the mutant K80R (Fig. 5A).
A
B
C
SUMOylation of HDGF K. Thakar et al.
1414 FEBS Journal 275 (2008) 1411–1426 ª2008 The Authors Journal compilation ª2008 FEBS
Fig. 1. HDGF is SUMOylated in mammalian cells. COS-7 cells transfected with plasmids coding for the indicated proteins were lysed and
analysed by SDS ⁄PAGE and western blotting with HDGF specific, Strep-tag specific, SUMO-1 specific, EGFP specific and Ubc9 specific anti-
bodies as indicated. (A) COS-7 cells were transfected with plasmids coding for the expression of HDGF untagged (wt), alone or together
with EGFP-SUMO-1. (B) COS-7 cells were transfected with plasmids coding for the expression of HDGF untagged (wt), alone or together
with EGFP-SUMO-1 and were lysed in TNE buffer either in the presence or absence of N-ethylmaleimide and IAA (20 mMeach). In the
absence of N-ethylmaleimide and IAA, the higher molecular weight band starting to disappear after 15 min of incubation of the cleared cell
lysate on ice. (C) COS-7 cells were transfected with empty plasmids (mock) or plasmids coding for HDGFStrep-tag (wt), together with
EGFP-SUMO-1, EGFP-SUMO-1DGG, HA-Ubc9, or EGFP as indicated. All cell lysates were treated with StrepTactinbeads for specific pre-
cipitation of HDGF and SUMO-1 conjugated HDGF (see Experimental procedures). SUMOylated HDGF could only be detected in the eluates
of co-expressed HDGFStrep-tag and EGFP-SUMO-1wt with or without overexpression of Ubc-9 using antibodies directed against the Strep-
tag or SUMO-1 as indicated.
Fig. 2. Amino acid alignment of HDGF from different mammalian species. Sequence alignment of HDGF from human (hHDGF, Homo sapi-
ens, Genbank accession CAI16347), chimpanzee (cHDGF, Pan troglodytes NCB accession XP_513894), bovine (bHDGF, Bos taurus, NCB
accession CAB40626), murine (mHDGF, Mus musculus, NCB accession BAB30979) and rat (rHDGF, Rattus norvegicus, NCB accession
AAL47132) origin, respectively. Gaps introduced to generate this alignment are indicated by dashes. Lysine residues within potential
SUMOylation motifs are highlighted in gray. K80 in the identified nonconsensus SUMOylation motif RK
80
GF is boxed. The PWWP motif is
indicated by a black background. Sequences for NLS1 and 2 are underlined. The numbers and letters within the secondary structure ele-
ments of the hath region (amino acids 1–98) are shown as arrows (b-strands) and rectangles (a-helices) according to PDB file 2B8A. Amino
acid residues for each sequence are numbered from the initiation methionine.
K. Thakar et al. SUMOylation of HDGF
FEBS Journal 275 (2008) 1411–1426 ª2008 The Authors Journal compilation ª2008 FEBS 1415